Seagrass collapse due to synergistic stressors is not
1anticipated by phenological changes
2Giulia Ceccherelli1*,Silvia Oliva1, Stefania Pinna1, Luigi Piazzi1, Gabriele Procaccini2, 3
Lazaro Marin-Guirao2, Emanuela Dattolo2, Roberto Gallia2, Gabriella La Manna1,3, Paola 4
Gennaro4, Monya M. Costa5, Isabel Barrote5, João Silva5, Fabio Bulleri6 5
6 7
1Dipartimento di Scienze della Natura e del Territorio, Polo Bionaturalistico, University of 8
Sassari, via Piandanna 4, 07100 Sassari, Italy 9
2Stazione Zoologica Anton Dohrn, Villa Comunale, Napoli, Italy 10
3MareTerra – Environmental Research and Conservation, Regione Sa Londra 9, Alghero 11
(SS), Italy 12
4Italian National Institute for Environmental Protection and Research (ISPRA), via di Castel 13
Romano 100, Roma, Italy 14
15
5CCMAR - Centre of Marine Sciences, University of Algarve, Faro, Portugal 16
6Department of Biology, University of Pisa, via Derna 1, Pisa 17 18 *corresponding author 19 GIULIA CECCHERELLI 20
Dipartimento di Scienze della Natura e del Territorio, 21
Polo Bionaturalistico, 22
University of Sassari, 23
via Piandanna 4, 07100 Sassari, Italy 24
phone: +39-079-228643 25
e-mail address: cecche@uniss.it 26
27
Author contributions: GC, SO, and LP conceived and designed the experiments. GC, SO, 28
SP and LP performed the experiments. GP, LMG, ED, RG, PG, MMC, IB, JS analysed the 29
sampled material, FB analysed the data. GC and FB led the writing of the ms. All authors 30
contributed critically to the drafts and gave final approval for publication. 31
running title: Synergistic stressors cause seagrass collapse 32
33
34
Abstract 35
Seagrasses are globally declining and often their loss is due to synergies among 36
stressors. We investigated the interactive effects of eutrophication and burial on the 37
Mediterranean seagrass, Posidonia oceanica. A field experiment was conducted to 38
estimate whether shoot survival depends on the interactive effects of three levels of 39
intensity of both stressors and to identify early changes in plants (i.e. morphological, 40
physiological and biochemical, and expression of stress-related genes) that may 41
serve to detect signals of imminent shoot density collapse. Sediment burial and 42
nutrient enrichment produced interactive effects on P. oceanica shoot survival, as 43
high nutrient levels had the potential to accelerate the regression of the seagrass 44
exposed to high burial (HB). After 11 weeks, HB in combination with either high or 45
medium nutrient enrichment, caused a shoot loss of about 60%. Changes in 46
morphology were poor predictors of the seagrass decline. Likewise, few 47
biochemical variables were associated with P. oceanica survival (the phenolics, 48
ORAC and leaf 34S). By contrast, the expression of target genes had the highest
49
correlation with plant survival: photosynthetic genes (ATPa, psbD and psbA) were 50
upregulated in response to high burial, while carbon metabolism genes (CA-chl, 51
PGK and GADPH) were down-regulated. Therefore, die-offs due to high 52
sedimentation rate in eutrophic areas can only be anticipated by altered expression 53
of stress-related genes that may warn the imminent seagrass collapse. 54
Management of local stressors, such as nutrient pollution, may enhance seagrass 55
resilience in the face of the intensification of extreme climate events, such as floods. 56
Key words: burial, early warnings, eutrophication, multiple-stressors, Posidonia oceanica. 57
1. Introduction 59
Transitions between natural systems with radically different properties can occur 60
abruptly. Important examples can be found in ecology (e.g., lake eutrophication and coral 61
reef collapses), where regime shifts have consequences that are often irreversible 62
(Bellwood et al. 2004, Carpenter 2011, Perry and Morgan 2017). Predicting and 63
anticipating catastrophic regime shifts represents a timely objective to improve our ability 64
to preserve biodiversity and ecosystem functioning in the face of escalating anthropogenic 65
pressures (Boettiger and Hastings 2013). Among all systems, the coastal marine are 66
experiencing a wide range of human-induced alterations, the magnitude of which 67
increases with the local density of human populations. Generally, the various 68
environmental stressors do not act in isolation (Crain et al. 2008); rather, the effects of 69
individual stressors interact to generate cumulative impacts that can be greater or smaller 70
than the sum of their individual impacts (i.e. synergistic or antagonistic effects). 71
Nevertheless, predicting the effects of combinations of stressors is particularly challenging 72
because mechanisms underpinning impact are rarely elucidated and either threshold or 73
nonlinear responses to stressors remain unknown (Griffen et al. 2016). 74
The effects of multiple stressors on slow-growing, habitat-forming species, such as 75
certain seagrasses, are particularly threatening. Seagrass meadows are among the most 76
important and productive coastal systems (Costanza et al. 1997). They provide key 77
ecological services, including nursery grounds, habitat (for a review see Heck et al., 2003), 78
organic carbon production and export, nutrient cycling, sediment stabilization, trophic 79
transfer to adjacent habitats (Hemminga and Duarte 2000, Larkum et al. 2006) and coastal 80
protection from erosion (Fonseca and Cahalan 1992, Fonseca and Koehl 2006). 81
Nonetheless, they are threatened by the rapid environmental changes caused by the 82
expansion of coastal human populations. Rapid, large-scale seagrass loss over relatively 83
short temporal scales has been reported throughout the world (Bulthuis 1983, Orth and 84
Moore 1983, Fourqurean and Robblee 1999, Marbà et al. 2005, Walker et al. 2006). 85
Stressors, such as sediments and nutrients inputs from terrestrial runoff, physical 86
disturbance (e.g. trawling, anchoring), invasive species, disease, aquaculture, overgrazing, 87
algal blooms and global warming, have been shown to cause seagrass declines at scales 88
ranging from square meters to hundreds of square kilometres (e.g. Munkes 2005, Orth et 89
al. 2006, Williams, 2007, Waycott et al. 2009, Bockelmann et al. 2011, Giakoumi et al.
90
2015). Overall, enhanced nutrients loading and sedimentation rates are likely the most 91
common and significant causes of seagrass decline (Unsworth et al. 2015). Indeed, the 92
current expansion of fish farming and other aquaculture practices (e.g., shellfish culture) 93
can have serious consequences on local populations of seagrasses through increased 94
deposition of organic matter and nutrients (Marbà et al. 2006). While eutrophication is 95
considered as the main cause of seagrass loss at a regional scale, burial of plants due to 96
anthropogenic-increased sedimentation or natural extreme events like storms or floods, 97
has been identified as an important cause for local die-offs (Short and Wyllie-Echeverria 98
1996, Erftemeijer and Lewis 2006, Orth et al. 2006, Cabaço et al. 2008, Cabaço and 99
Santos 2014). 100
Posidonia oceanica (L.) Delile is a slow-growing seagrass, endemic in the
101
Mediterranean and experiencing a widespread decline throughout the basin (Telesca et al. 102
2015). The regression of P. oceanica beds and the consequent expansion of alternative 103
habitats (e.g. algal turfs or dead seagrass rhizomes, generally referred to as “dead matte”) 104
is particularly common in the proximity of urban areas (Montefalcone et al. 2009, 105
Tamburello et al. 2012). In addition to enhanced nutrient loading, P. oceanica meadows 106
are exposed to increased sedimentation rates as a consequence of beach nourishment, 107
dredging of waterways, shoreline armouring and severe climatic events (i.e. storms and 108
floods). Correlative and experimental studies have assessed the effects of both 109
eutrophication (Alcoverro et al. 1997, Delgado et al. 1999, Ruiz et al. 2001, Holmer et al. 110
2008) and burial on P. oceanica (Manzanera et al. 2011, Gera et al. 2014), but how 111
organic load levels change the effects of burial is yet to be explored. Within this context, 112
the identification of early warning signals for drastic declines would be a valuable tool for 113
the management of seagrass meadows (McMahon et al. 2013, Macreadie et al. 2014, 114
Roca et al. 2016). This goal has been pursued for pressing disturbances through 115
correlative approaches (e.g., van Katwijk et al. 2011) that do not, however, allow 116
estimating signs of imminent collapses. 117
Here, we investigate the interactive effects of eutrophication and burial on P. 118
oceanica. A field experiment was conducted 1) to estimate whether shoot survival
119
depends on the interactive effects of three levels of intensity of both stressors, using a full 120
factorial design, and 2) to identify early changes in plant attributes (i.e. 121
morphological/growth, physiological/biochemical, and expression of stress-related genes) 122
that may serve as signals of imminent shoot density collapse (early warnings of 123
degradation). In general, response time and sensitivity to stressors vary with the type of 124
variable examined; thus, understanding how sensitivity to stressors may change according 125
to the level of biological organization is essential to rationalise the choice of indicators of 126
impending seagrass decline and to design monitoring programmes (Roca et al. 2016). 127
Indicator specificity is expected to increase when moving towards lower levels of biological 128
organisation (sensu, Whitham et al. 2006), from the structural metrics to specific 129
physiological and molecular indicators (Adams and Greeley, 2000). Whether this general 130
rule holds for P. oceanica remains utterly unexplored. 131
2. Materials and Methods 132
2.1. Study site 133
The study was carried out in a shallow (5-8 m deep) continuous P. oceanica 134
meadow, on the north-west coast of Sardinia (40°34.1 N, 09°8.5 E). At this site, P. 135
oceanica canopy structure in the inner meadow is well preserved (shoot density mean±SE
136
= 699.4±38.6 m-2, n=28; canopy height mean±SE = 63.92±1.94 cm, n=35). At the site, the 137
grazing sea urchin, Paracentrotus lividus, is present at low densities, whilst juveniles of the 138
herbivore fish, Sarpa salpa, are common (GC, personal observations). 139
2.2. Experimental design and set up 140
The experiment started on the 29th of April 2015 and lasted until P. oceanica shoot 141
mortality exceeded 60% in some treatments (i.e. 11 weeks, see Results). The hypotheses 142
were tested by running a fully-factorial experiment of sediment burial (high, medium and 143
control) and nutrient enhancement (high, medium and ambient) treatments. Twenty-seven 144
circular patches (hereafter referred to as experimental units) were randomly selected 145
across the seagrass meadow, at least 3 m apart one from another. Within each unit, a 146
PVC cylinder (40 cm in height and diameter) was inserted about 10 cm deep into the 147
bottom, thus leaving about 30 cm of the cylinder above the sediment (Fig. S1). Each 148
cylinder enclosed between 81 and 109 P. oceanica shoots. 149
High, medium and control burial (HB, MB and CB) were obtained by adding 12.5 L, 150
5.0 L and 0 L of sediment (corresponding to a layer 10 cm, 4 cm and 0 cm tick), 151
respectively accordingly with Manzanera et al. (2011). We used washed sand for 152
playgrounds in our experiment, as this sand is similar in grain size (coarse sand, 0-1 mm) 153
and mineralogy (carbonate) to sediment at the study site. This ensured that all 154
experimental units were treated with the sand characterized by the same granulometry, 155
devoid of fauna and low in organic content. Depending on the treatment, P. oceanica 156
plants were buried with 4 cm of sediment over the ligula in HB, at the ligula height in MB, 157
and unburied in CB. When scuba divers filled experimental units with sand, care was taken 158
not to damage the leaves and to keep the plants upright during the process. 159
Also, high, medium and ambient nutrients (HN, MN and AN) were obtained by adding 160
80 g, 40 g and 0 g of homogenized fish fodder (protein 56.66%, fat 24.88%, cellulose 161
5.53%, ash 9.4%, phosphorous 1.35%, calcium 1.73% and sodium 0.46%) to the sediment 162
in each unit. Treatment levels were decided on the basis of nutrient release from offshore
163
fish farms (Garcìa-Sanz et al. 2010). Fish fodder was used to simulate effects of
164
eutrophication related to land and coastal use change (from deforestation to aquaculture), 165
which often increases the organic matter and nutrients input into coastal sediments. The 166
use of fish fodder not only increases ammonium levels through mineralization, but also 167
fuels the production of sulfide (Burkholder et al. 2007). Sulfide is a strongly phytotoxic 168
compound, as it blocks the activity of cytochrome oxidase and other metal containing 169
enzymes, which may lead to massive seagrass die-offs (Govers et al. 2014). 170
Each unit was randomly assigned to one of the six combinations of treatments (n=3). 171
Three extra cylinders with large holes and not exposed to either sediment or fish fodder 172
addition were used as procedural controls. The outer edge of all units was not parted off 173
and a trans-cylinder clonal integration between shoots was not impeded with the aim not 174
to impose further stress on shoots. The effects of the experimental conditions on shoot 175
survival were evaluated every few weeks (see section 2.3.2) with the aim of catching the 176
shoot mortality of about 20% and 60% in the harshest treatments that had corresponded 177
to week 3 and 11. Then, to identify indicators of mortality, shoot survival at week 11 was 178
related with morphological, physiological/biochemical variables and expression of stress-179
related genes, estimated at week 3. 180
2.3. Data collection and analyses 181
2.3.1. Assessment of trophic enrichment 182
In order to estimate concentrations of inorganic nutrients, samples were taken from 183
the water column at two dates chosen at random during the experiment (week 7 and 11). 184
At each date, two replicate water samples were taken in each unit using a 125-mL sterile 185
bottle about 5 cm above P. oceanica rhizomes. Samples were shaken and then filtered 186
(0.45 μm mesh size) as soon as they were brought to the surface. Samples were frozen in 187
liquid nitrogen for transportation to the laboratory, where concentrations of ammonia, 188
nitrate, nitrite and ortophosphate were determined using a continuous-flow AA3 189
AutoAnalyzer (Bran-Luebbe) and expressed in μmol l–1. Concentrations of dissolved 190
inorganic nitrogen (DIN, ammonia+nitrate+nitrite) and phosphorus (DIP as P-PO43-) were 191
analysed by means of two one-way analyses of variance (ANOVA) testing the effect of 192
Nutrient enrichment (3 levels, HN, MN and AN) on the pooled data taken in each unit 193
(n=18). Cochran's C-test was used before each analysis to check for homogeneity of 194
variance, and data were transformed when necessary. SNK test was used for a posteriori 195
means comparisons (Underwood 1997). 196
2.3.2. P. oceanica survival 197
P. oceanica shoot density was counted at week 0, 3, 7, 9 and 11 to estimate shoot
198
survival (assessed as the percentage of shoots with leaves) through time. Shoot survival 199
(%) after 3 and 11 weeks was analysed by means of two 2-way analyses of variance 200
(ANOVA), including the factors Burial (3 levels, HB, MB and CB, fixed) and Nutrient 201
addition (3 levels, HN, MN and AN) both fixed and orthogonal (n=3). Cochran's C-test was 202
used before each analysis to check for homogeneity of variance and data were 203
transformed when necessary (Underwood 1997). 204
To evaluate the effects of procedural controls (PC), P. oceanica shoot survival at 205
week 3 and 11 was also analysed by two one-way ANOVAs (PC vs. CBAN, n=3). 206
Cochran's C-test and SNK test were run as explained above. 207
2.3.3. P. oceanica morphological/growth and physiological/biochemical variables 208
Six morphological variables (epiphyte load, % of leaves with necrosis, maximum leaf 209
length, mean leaf length, number of leaves, shoot biomass), leaf growth rate, and eleven 210
physiological/biochemical variables (leaf N, C and S content and corresponding isotopic 211
signature 15N, 13C, 34S, the antioxidant capacity through the oxygen radical absorbance 212
capacity, ORAC, Trolox equivalent antioxidant capacity, TEAC and phenolics) (Table 1) 213
were estimated after 3 weeks, when shoot mortality was still inconspicuous (see Results). 214
Selection of these variables was based on the review made by Roca et al. (2016). 215
Among the morphological variables, epiphyte load was estimated as the weight of the 216
epiphytes obtained from scraping the shoots with a razor blade (referred to the total weight 217
of scraped leaves). The incidence of leaf necrosis was estimated as the percentage of 218
leaves showing marks of necrosis (black areas or spots). Leaf growth rate was assessed 219
using a modified leaf punching technique. At the beginning of the experiment, three shoots 220
per unit were marked by punching a hole just above the leaf base-leaf blade junction of the 221
outermost leaf with a hypodermic needle, and tagged with a plastic cable tie. After 20 222
days, the punched shoots were collected and the length of the newly produced tissue in 223
each shoot measured. 224
Among the physiological/biochemical variables, ORAC, TEAC and phenolics were 225
extracted in frozen leaf tissue samples as described in Costa et al. (2005) and estimated 226
following Huang et al (2002), Re et al. (1999), and Folin-Ciocalteu method (Booker and 227
Miller 1998, Migliore et al. 2007), respectively. Leaf N, C, and S (%) were determined in 228
samples of ca. 3.5 mg of dried, finely ground and homogenized material from each unit. 229
Leaf N, C and S isotopic signature, was analysed through elemental analyser combustion 230
for continuous flow isotope ratio mass spectroscopy. 231
Effects of Burial (3 levels, HB, MB and CB, fixed) and Nutrients (3 levels, HN, MN 232
and AN) on morphological/growth and physiological/biochemical variables (n=3) at week 3 233
were analysed by means of two 2-way analyses of variance (ANOVA). Cochran's C-test 234
was used before each analysis to check for homogeneity of variance (Underwood 1997). 235
2.3.4. P. oceanica expression of stress-related genes 236
The expression of ten stress-related genes (GADPH, SHSP, HSP90, PGK, CAB-151, 237
psbD, psbA, CA-chl, rbcl, and ATPa), (Table 1) were estimated after 3 weeks. At this aim 238
three shoots for each experimental unit were collected after 3 weeks of treatment. A 4 cm 239
leaf segment from the youngest fully mature leaves of each shoot (usually the second-rank 240
leaf) was collected and rapidly cleaned from epiphytes with a razor blade, towel-dried and 241
immediately stored in RNAlater® tissue collection solution (Ambion, Life Technologies). 242
Samples were then transported to the laboratory, preserved one night at 4 ◦C and stored 243
at −20 ◦C until RNA extraction. 244
After total RNA extraction (Mazzuca et al. 2013), RNA quantity and purity were 245
assessed by Nanodrop (ND-1000 UV-Vis spectrophotometer; NanoDrop Technologies) 246
and 1% agarose gel electrophoresis. Average Abs260/280 nm and Abs260/230 nm ratios
247
(2.0 and 1.8, respectively) have indicated the absence of protein and solvent
248
contaminations, while gel electrophoresis has showed intact RNA, with sharp ribosomal
249
bands. Total RNA (500 ng) was reverse-transcribed in complementary DNA (cDNA) with
250
the iScript™ cDNA Synthesis Kit (Bio-Rad) using the GeneAmp PCR System 9700 (Perkin 251
Elmer). 252
The target genes were selected on the basis of the two key events that mandatorily 253
occur in the light and dark reactions of photosynthesis and can be hampered by biotic and
254
abiotic stressful conditions, reducing plant productivity and survival (Ashraf and Harris,
255
2013). In light reactions, light energy is harvested by antenna pigments, channelled to
256
photosystem II (PSII) to produce photochemistry and converted into chemical energy (i.e. 257
ATP) and reducing power (i.e. NAPDH) by the flow of electrons along the electron 258
transport chain. Particularly, we have selected and analyzed genes coding for a putative 259
fundamental protein of the photosynthetic antenna complex (CAB-151): the two PSII core 260
proteins D1 and D2 (psbA and psbD) and a putative chloroplastic ATP synthase subunit 261
alpha (ATPa). In dark reactions, CO2 is fixed into carbohydrates in the Calvin-Benson 262
cycle by using ATP and NADPH produced in the light reactions. In relation to this cycle, we 263
have selected several putative genes encoding key enzymes of the plant carbon 264
metabolism: the large subunit of the RuBisCO (rbcl), directly involved in CO2 fixation, and 265
a chloroplast carbonic anhydrase (CA-chl) that catalyzes the conversion of HCO -3 into 266
CO2. We also selected two major enzymes involved in glycolysis, glyceraldehyde-3-267
phosphate dehydrogenase (GAPDH) and phosphoglycerate kinase (PGK), for obtaining 268
energy and carbon molecules from glucose. Finally, two general stress genes were also 269
included in the analysis as they are ubiquitous heat stress proteins in environmental 270
stress: the molecular chaperone (HSP90) and one putative small heat shock protein 271
(SHSP). All selected genes were already investigated in other studies (Table S1). Primers 272
sequences and GenBank Accession Numbers are reported in the reference in which the 273
genes have been selected for the first time. The expression stability of a set of three 274
putative reference genes already tested in P. oceanica (i.e. EF1A, 18S and L23; Serra et 275
al. 2012) was evaluated by running GeNorm (Vandesompele et al. 2002) and Normfinder
276
(Andersen et al. 2004). According to stability analysis, two genes, namely 18S and L23 277
(Fig. S2), were used to normalize gene expression data. 278
After normalizing by each primer efficiency (all >0.92% and R2>0.96), relative 279
treatment gene expression values were calculated as the negative differences in cycles to 280
cross the threshold value (-ΔCT) between the reference genes and the respective target 281
genes (-CT = CTreference - CTtarget). Subsequently, fold expression changes were 282
calculated for graphical purposes according to the equation: fold expression change = ± 283
2(│(-ΔCTtreatment)-(-ΔCTcontrol)│). 284
Effects of Burial (3 levels, HB, MB and CB, fixed) and Nutrients (3 levels, HN, MN 285
and AN) on the relative expression of target genes (-CT values; n=3) were analysed by 286
means of 2-way analyses of variance (ANOVA). Cochran's C-test was used before each 287
analysis to check for homogeneity of variance (Underwood 1997). 288
2.3.5. Assessment of the predictive variables 289
In order to identify predictors of shoot survival of P. oceanica under different 290
combinations of burial and nutrients enhancement levels, separate multiple regressions 291
were run between each of the four groups of response variables included in the study 292
(morphological/growth, antioxidant, isotopes, and gene expression) and shoot survival. 293
More specifically, values of explanatory variables measured at week 3 were used as 294
predictors of shoot survival at week 11. Collinearity among covariates was assessed by 295
means of Variance Inflation Factor (VIF) procedures. Covariates with highest VIF values, 296
calculated using the R car package, were sequentially dropped from the model, until all 297
VIF values were smaller than 3, as recommended by Zuur et al. (2010). Linearity and 298
homogeneity of variances was visually checked by means of residual plots. Log 299
transformation of data was effective in enhancing homogeneity of variances. For each 300
group of variables, the best-fit model was selected by means of a stepwise procedure 301
(both directions) using the stepAIC function from the R MASS package (Venables and 302
Ripley 2002). This function selects the best model based on the Akaike Information 303
Criteria. The relationship between plant survival and predictive variables retained by the 304
different best-fit models were visualized by means of partial regression plots, using the 305
function avPlots from the car package. The relative importance (as percentage) and 306
bootstrap confidence intervals of the explanatory variables retained in the best fit models 307
were assessed by means of the Lindemann-Merenda-Gold (lmg) method for calculating 308
sequentially weighted partial R2 (Lindeman et al. 1980), using the R “relaimpo” package 309
(Gromping 2006). This method calculates an average coefficient of partial determination 310
for each model permutation using the individual contribution of each explanatory variable. 311
3. Results 312
3.1. Inorganic nutrients and P. oceanica survival 313
The addition of fish fodder to sediments resulted in a significant increase of DIN 314
concentration in the water column above rhizomes (F2,51=10.93 p=0.0001; Fig. 1); as 315
shown by the SNK tests, the increment in DIN was proportional to the amount of fodder 316
added (HN>MN>AN). By contrast, there were no significant differences in DIP among 317
treatments (F2,51=3.08 p=0.0547). 318
The comparison between CBAN and PC shows that there was no artefact of PVC 319
cylinders on shoot mortality at both week 3 (F1,4=0.25 p=0.6430) and week 11 (F1,4=0.04 320
p=0.8554) (Fig. 2). P. oceanica shoot survival was significantly affected by sand burial, as 321
about the 25% of shoots died by week 3 in the HB units (Fig. 2 and Table 2, HB<MB=CB). 322
Burial was the only stressor affecting mortality of the seagrass in the short-term (3 weeks). 323
After that, shoot survival decreased through time and, after 11 weeks, it was regulated by 324
the interactive effects of nutrient addition and burial (Table 2; Fig. 2). In units exposed to 325
high burial, shoot survival under high and intermediate nutrients addition was lower than at 326
ambient nutrient levels (HBHN=HBMN<HBAN). In contrast, shoot survival was not affected 327
by nutrients levels in units exposed to either medium or ambient burial (SNK tests in Table 328
2). At high nutrients levels, shoot survival was lower in units exposed to high than medium 329
or ambient burial levels (HBHN<MBHN=CBHN). At medium nutrients levels, the survival 330
decreased with increasing severity of burial (HBMN<MBMN<CBMN), while at ambient 331
nutrient levels, there was no difference among burial treatments (Fig. 2 and Table 2, 332
CBAN=MBAN=HBAN). 333
Synergistic effects of both high and medium nutrients addition and high burial were 334
further assessed by comparing the average response of P. oceanica shoot survival (%), 335
both at each stressor level and at each combination of stressor levels, with those obtained 336
for the theoretical additive response of stressors levels in combination (Fig. 3). In fact, for 337
the average response given at high nutrient addition (HN), medium nutrient addition (MN), 338
and high burial (HB) when applied individually, the cumulative effects under additive 339
conditions (HN+HB and MN+HB) on shoot survival would have been about two fold higher 340
than that observed (Fig. 3). The difference in shoot survival between the average 341
experimental evidence and the theoretical additive estimates provides evidence for the 342
synergism between nutrient (high and medium) and burial (high) stressors. 343
3.2. P. oceanica predictive variables 344
After three weeks since treatments, nutrient addition changed epiphyte load, necrosis 345
and leaf growth rate, among morphological/growth variables, phenolics and the antioxidant 346
response, ORAC and TEAC, among the physiological variables. However, there was no 347
significant effect of nutrient enrichment on isotopes or gene expression, except for the 348
stress gene HSP90, which experienced a slight inactivation (Fig. 4, 5, 6 and Table 3). In 349
contrast, burial significantly increased epiphyte load and affected gene expression of 350
GADPH, PGK, psbA, CA-chl and ATPa (Table 3). 351
The only variables that responded to the interactive effects of nutrient addition and 352
burial were the percentage of leaf necrosis and phenolics content, being one the mirror 353
pattern of the other (Table 3): the first was significantly greater at high nutrient levels when 354
plants were exposed to high burial (HN>MN=AN) and at high burial levels if plants were at 355
ambient nutrients levels (HB>MB=CB), while the latter was smaller at high nutrients levels 356
condition only when plants were exposed to HB (HN<MN=AN), and in high buried plants 357
only at HN (HB<MB=CB). 358
The multiple regressions conducted with genetic, isotope, antioxidant and 359
morphological covariates explained the 60.6%, 33.2%, 36.4% and 27.8% (Adjusted R2
360
values) of the variance in shoot survival, respectively (Table 4). After simplification through 361
the step-wise procedure, the best fit model at the level of gene expression included two 362
explanatory variables, CA-chl and ATPa, although only the effects of the former was 363
significant, contributing nearly for 84% of the variation in shoot mortality explained by the 364
model (Table 5; Fig. 6). The positive estimate of CA-chl suggests that overexpression of 365
this gene at week 3 is positively correlated with shoot survival at week 11 (Fig. S3). CA-chl 366
is directly related with carbon metabolism (correlated to PGK and GADPH) and was, in 367
fact, down-regulated in treatments, such as HBHN and HBMN (SNK test, Table 6), where 368
P. oceanica had the lowest survival. ATPa is a gene involved in the photosynthesis
369
(correlated to psbA and psbD) and was up-regulated in high burial treatments (Fig. 6 and 370
Table 6). 371
Four explanatory variables were retained in the best fit isotope model and only one of 372
these, namely leaf 34S, was significantly related with shoot survival at 11 weeks and 373
accounted for about 50% of the total variation in shoot survival explained by the model 374
(Table 4). However, although leaf 34S was positively (Fig. S4) related with shoot survival, 375
the ANOVA did not identify any significant effect neither of nutrient addition nor of burial on 376
leaf 34S content (Table 3). The best fit antioxidant model retained two variables, ORAC 377
and phenolic content, both of which were significantly correlated with shoot survival. 378
Phenolic contents accounted for a greater proportion (~63%) than ORAC (~37%) of the 379
total variability in shoot mortality explained by the model (Table 4). Coefficient estimates 380
were negative for ORAC and positive for phenolics (Fig. S5). 381
The morphological/growth best fit model retained 3 variables, epiphyte load, shoot 382
biomass and growth (Table 4). The former variable accounted for most of the total 383
variation in shoot survival explained by the model (~77%, Table 5). The negative estimate 384
of the relationship coefficients (Table 4) indicates that increasing levels of epiphytic 385
loading at week 3 were associated with lower shoot survival at week 11 (Fig. S6). Plants 386
that had grown more in the first three weeks had greater chances of be alive at week 11. 387
DISCUSSION 388
4.1. Synergistic effects of nutrients and burial on P. oceanica mortality 389
P. oceanica shoot survival was affected by synergistic effects of burial and nutrient
390
addition over a short time. High sediment load (HB), corresponding to the complete burial 391
of meristems, increased seagrass mortality independently of the organic load in just three 392
weeks. Apparently, HB was the only level strong enough to cause rapid seagrass 393
mortality. 394
Over a longer term, burial and nutrient enrichment had interactive, negative effects 395
on P. oceanica survival. In particular, high nutrients levels (HN) had the potential to 396
accelerate the regression of the seagrass subjected to high burial. The use of a gradient of 397
both stressors, manipulated in a crossed design, has also allowed detecting the rapid 398
threshold responses to their combinations. Thus, by week 11, HB in combination with 399
either HN or MN, caused a shoot loss of about 60%. In addition, the slow decrease in 400
shoot survival under high burial without organic loading (HBAN), suggests that mortality at 401
HB was determined by the level of nutrient load, from the fastest at HN to the slowest at 402
AN. 403
Total shoot mortality did not occur within 11 weeks since the start of the experiment, 404
even under the most stressful conditions and it is unknown whether it would have occurred 405
on a longer term. However, several mechanisms may underpin the survival of some 406
shoots within experimental plots. First, the clonal integration with neighbouring shoots 407
immediately outside the unit edge: continuity between shoots was not prevented to avoid 408
further stress that would have biased the response of the plants. Thus, transfer of 409
metabolites (e.g. carbohydrates) from non-treated plants may have sustained the survival 410
of plants experimentally exposed to stressful conditions. Second, burial effect might have 411
not been homogeneous among shoots within each experimental unit, as rhizome height 412
varied among them and samples were only three-replicated. Third, individual response to 413
stress can differ. Differences in resilience can be related to individual characteristics (e.g. 414
age) of single ramets or to genetic peculiarities of single genets. More resilient genotypes 415
could respond better to the synergistic action of the applied stressors (Hughes and 416
Randall 2004). 417
4.2. Predictive variables of P. oceanica collapse 418
The protraction of the experiment for 11 weeks allowed detecting widespread 419
mortality in experimental units exposed to HBHN and HBMN. Thus, besides the 420
identification of lethal effects of burial and eutrophication on P. oceanica, our study 421
enabled the identification of the plant attributes (morphological/growth, 422
physiological/biochemical and transcriptomic) at week 3, prior to mortality, related to 423
survival at week 11. 424
4.2.1. Morphological/growth variables 425
Except for the increase in epiphyte load, changes in morphological variables and 426
growth rate were little effective in predicting seagrass loss (Table 6). An increase in 427
epiphyte biomass, especially macroalgae, is a common response to eutrophic conditions 428
(Piazzi et al. 2016) and the ratio between epiphyte and leaf biomass in P. oceanica is one 429
of the descriptors used to assess the ecological quality of Mediterranean water bodies 430
under the European Water Framework Directive (Gobert et al. 2009, Oliva et al. 2012). 431
Our study shows that the epiphyte load can be used also as an indicator of burial stress, 432
as probably a consequence of lower phenolics content in the buried plants (Costa et al. 433
2015). Furthermore, signals of imminent mortality can also be gained by the percent of 434
leaves with necrosis; leaf necrosis has been previously suggested to increase with 435
eutrophication (Roca et al. 2016). Here, a higher proportion of leaves exposed to both high 436
burial and both high and medium nutrient addition had necrotic spots which were not 437
uniformly distributed along the blades, being mostly concentrated on the basal part of the 438
leaf. Finally, leaf growth was accelerated by nutrient addition, but was not related to burial, 439
suggesting that this variable is inadequate for predicting imminent seagrass degradation 440
due to the combination of stressors. 441
4.2.2. Physiological/biochemical variables 442
Based on the multiple regressions, the biochemical/physiological variables 443
associated with the P. oceanica survival, were the phenolic content, ORAC and 34S.
444
Among these, only phenolics are likely to be a useful predictor (positive coefficient), since 445
differences in their concentration were also related to the interactive effect of nutrient 446
addition and burial (Table 6). Phenolics can have multiple biological functions mainly 447
related to the reproductive strategy, adaptation and survival to environmental 448
disturbances, antimicrobial and antifouling properties. Their deposition as lignin in cell 449
walls increases their mechanical strength and improves plant response against pathogens 450
and wounding. Variations in phenolics content have been observed in P. oceanica as a 451
response to changes in water quality and when competing with invasive species (Pergent 452
et al. 2008, Migliore et al. 2007, Rotini et al. 2013).
453
Furthermore, ORAC, which reflects the ability of the plant metabolism to scavenge 454
oxygen reactive species (ROS, Mittler, 2002) through hydrogen atom donation (both 455
enzymatically and non-enzymatically), was negatively associated with P. oceanica 456
survival, and shoots with higher ORAC will have higher probability of mortality for the high 457
concentrations of ROS. However, it only responded to nutrient addition (Table 6). 458
Furthermore, the 34S content (positive coefficient) would suggest that the presence of
459
sulphide intrusion in leaf tissue, lower leaf 34S signature, could predict greater shoot 460
survival (Table S4). However, this isotopic signal did not respond uniformly to any 461
treatment, probably due to the sediment type (i.e. coarse-carbonate sediments, Oliva 462
2012): indeed, the relationship between sediment sulfide and the 34S of P. oceanica
463
tissues is known to be rather complex being controlled both by the sediment sulfide 464
concentrations and the oxygen status of the plants (Borum et al. 2005). Because of the 465
high variability, the use of 34S as an indicator for fish farm effects on P. oceanica has
466
already not been supported (Frederiksen et al. 2007). Therefore, among the 467
biochemical/physiological variables only results in leaf phenolics content are promising 468
enough to promote it as an indicator of nutrient and burial stressors of P. oceanica (Table 469
6). 470
4.2.3. Expression of selected genes 471
Gene expression of target genes had the highest correlation with shoot survival. 472
Photosynthetic genes (ATPa, psbD and psbA) were up-regulated in response to high 473
burial suggesting that high sediment load tends to increase the number of PSII and ATP 474
synthase, probably to compensate for high energy consumption. By contrast, carbon 475
metabolism genes (CA-chl, PGK and GADPH) were down-regulated, being associated 476
with plant survival. In particular, the experimental treatments significantly affected the 477
ability of plants to fix C and recycle ATP through glucolysis or gluconeogenesis, as 478
suggested by the significant down-regulation of PGK. This response is likely to result in a 479
long-term starvation of the plants, likely contributing to the high mortality observed in these 480
treatments. In other words, burial produced a strong effect on the carbon metabolism of 481
plants, reducing carbon fixation and the production of energy: buried plants invested 482
energy to increase the amount of PSII and ATP synthase to sustain the thylacoid proton 483
gradient necessary to produce more ATP. Further, because HSP90 is an ATP-dependent 484
molecular chaperone it is possible that a strong ATP reduction experienced by plants from 485
the most intense treatments resulted in lower expression levels of this gene. It is also 486
known that inhibition of HSP90 induced cell death in plants (Nishizawa-Yokoi et al. 2010; 487
Moshe et al. 2016) and, likely, the observed down-regulation is promoting and anticipating 488
plant mortality two months in advance. 489
4.3. Conclusions 490
This study has estimated that P. oceanica shoot survival strongly depends on the 491
interactive effects of levels of intensity of nutrients and burial and it has identified some 492
early changes in plant attributes. Within the whole set of possible predictors of seagrass 493
collapse, only a few were associated with shoot survival (with different contribution to the 494
total variability) and these did not necessarily correspond to those affected by the 495
treatments (Table 6). Furthermore, they were attributed to a different reliability level based 496
on the association to shoot survival. Overall, the evaluation of the candidate indicators of 497
P. oceanica regression for the consequences of high organic load and sediment burial has
498
produced some very promising evidence for the gene CA-chl, and secondly of ATPa, while 499
phenolics and epiphyte load are likely to be less reliable predictors. Thus, the sensitivity of 500
P. oceanica attributes to these stressors has increased as the level of biological
501
organization decreases (according to expectations, Roca et al. 2016), underlining the 502
importance of implementing monitoring protocols with biochemical and molecular 503
indicators. However, high-replicated studies that include different field conditions (e.g. 504
sediment type, wave exposure) are needed so that consistency of variables’ response 505
would be estimated and solid predictors incorporation promoted in any kind of 506
management program (i.e. assessment of ecosystem status, environmental quality, 507
impacts or the results of mitigation actions). 508
Nevertheless, no morphological change of shoot or leaf size should be expected after 509
a severe storm or flood that accumulates high loads of sediments, rich in organic matter, 510
over a P. oceanica bed; these attributes are likely to be reliable warnings only of pressing 511
disturbances over a longer temporal scale (Roca et al. 2016). Consequently, if other levels 512
of biological organization besides the morphological are not considered, the seagrass 513
collapse without early warning signals (Pace et al. 2015). Our results indicate that the 514
interaction between two of the main regional anthropogenic stressors in temperate coastal 515
ecosystems, eutrophication and high sediment loads (the deposition of organic detritus 516
likely brings with it increased sediments, Terrados et al. 1999), can trigger the fast collapse 517
of seagrass meadows without any major phenological change. 518
In the growing interest of identifying the type and role of interactions among local 519
anthropogenic stressors in driving habitat shifts in marine ecosystems (Russell and 520
Connell 2012), this controlled factorial experiment provides evidence that eutrophication 521
and burial have non-additive consequences on seagrass beds. The input of excess 522
nutrients (primarily nitrate and phosphate) to the marine environment is a global problem 523
associated with a range of human activities, but coastal eutrophication through organic 524
matter dispersal represents a further source of disturbance. In fact, the addition of a 525
detrital layer to a seagrass bed will not just result in increased dissolved inorganic nutrients 526
to the sediments, it will have a whole stimulatory impact upon the microbial, fungal and 527
detrital feeding community (Danovaro et al. 1994). A likely additional effect of increased 528
organic detritus is that of increasing the levels of sulphide stress within the sediments with 529
follow-up negative effects upon the seagrass (Marbà et al. 2006). Indirectly, our results 530
provide strong evidence suggesting that development of coastal areas and their 531
associated human activities will have major impacts on the seagrass meadows and such 532
information should be used to identify appropriate local management actions to halt the 533
global loss of seagrasses in favour of alternative habitats composed by either macroalgae 534
or anoxic mud (Unsworth et al. 2015). 535
Finally, local anthropogenic stressors are thought to negatively interact with global 536
climatic stressors resulting in decline of many habitats, such as coral reefs, macroalgal 537
forests, mangroves and seagrasses. These findings suggest that shifts from seagrass 538
systems to dead rhizomes (i.e. dead matte) in areas characterized by poor water quality, 539
may become more common under future scenarios of climate change (Garcia et al. 2013), 540
as frequency and intensity of storms and floods are expected to increase. 541
Because our study has demonstrated that the effects of nutrients and sediment burial 542
on seagrasses are synergistic, strategies for reducing nutrient levels, a widely advocated 543
strategy for seagrass conservation, should be pursued bearing in mind the need to control 544
also sedimentation rates. Local management of nutrient loading can represent a valid tool 545
for mitigating the impacts of global stressors on marine macrophytes (Falkenberg et al. 546
2010). Our study shows that reducing nutrient loading may be not sufficient to enhance the 547
resistance of seagrasses to global stressors, such as seawater warming (NOAA 2016). 548
Given the cumulative nature of human impacts and the large small-scale variation in life-549
traits occurring in coastal environments, one size fits all strategies are unlikely be 550
successful for sustaining the functioning of marine ecosystems in the face of climate 551
changes. In addition, our study makes a first step towards the identification of response 552
variables that may function as early warning signals of imminent seagrass collapse. It also 553
provides evidence for increased indicator specificity at lower levels of biological 554
organisation, promoting the need of implementing monitoring with molecular analyses. As 555
a final note of caution, the robustness of the response variables identified as most 556
promising must be assessed against variations in stressor intensity and background 557
abiotic and biotic conditions before their implementation in monitoring programs. 558
Acknowledgements 559
Analyses were performed at the Servizos de Apoio á Investigación-Universidade da 560
Coruña (Leaf N, Leaf C, and corresponding isotopic signatures), Iso-Analytical Ltd (Leaf S 561
and leaf 34S). We are grateful to Capo Caccia-Isola Piana MPA administration for 562
facilitating the samplings within their domains. This research was supported by MIUR, the 563
Italian Ministry of Education and Research, (TETRIS grant 2011-2015) and by Fondazione 564
di Sardegna (EIMS grant 2016-2019). PG acknowledges the support of S. Porrello and E. 565
Persia at ISPRA laboratories for water nutrient analysis. 566
567
LITERATURE CITED 568
Adams SM, Greeley MS (2000) Ecotoxicological Indicators of Water Quality: Using Multi-569
response Indicators to Assess the Health of Aquatic Ecosystems. Water, Air, and Soil 570
Pollution 123:103-115. 571
Alcoverro T, Romero J, Duarte CM, López NI (1997) Spatial and temporal variations in 572
nutrient limitation of seagrass Posidonia oceanica growth in the NW Mediterranean. 573
Marine Ecology Progress Series 146:155−161. 574
Andersen CL, Jensen JL, Orntoft TF (2004) Normalization of realtime quantitative reverse 575
transcription-PCR data: a model-based variance estimation approach to identify genes 576
suited for normalization, applied to bladder and colon cancer data sets. Cancer 577
Research 64:5245–5250. 578
Ashraf M, Harris PJC (2013) Photosynthesis under stressful environments: An overview.
579
Photosynthetica 51 (2):163-190.
580
Bellwood DR, Hughes TP, Folke C, Nyström M (2004) Confronting the coral reef crisis. 581
Nature 429:827–833. 582
Boettiger C, Hastings A (2013) From patterns to predictions. Nature 493:157-158. 583
Bockelmann AC, Beining K, Reusch TBH (2011) Widespread occurrence of endophytic 584
Labyrinthula spp. in northern European eelgrass Zostera marina beds. Marine Ecology
585
Progress Series 445:109-116. 586
Booker FL, Miller JE (1998) Phenylpropanoid metabolism and phenolic composition of 587
soybean [Glycine max (L.) Merr.] leaves following exposure to ozone. Journal of 588
Experimental Botany 49:1191-1202. 589
Borum J, Pedersen O, Greve TM, Frankovich TA, Zieman JC, Fourqurean JW, et al. 590
(2005) The potential role of plant oxygen and sulphide dynamics in die-off events of 591
the tropical seagrass, Thalassia testudinum. Journal of Ecology 93:148–158. 592
Bulthuis DA (1983) Effects of in situ light reduction on density and growth of the seagrass 593
Heterozostera tasmanica (Martens ex Aschers.) den Hartog in Western Port, Victoria,
594
Australia. Journal of Experimental Marine Biology and Ecology 67:91–103. 595
Burkholder JM, Tomasko DA, Touchette BW (2007) Seagrasses and eutrophication. 596
Journal of Experimental Marine Biology and Ecology 350:46–72. 597
Cabaço S, Santos R (2014) Human-induced changes of the seagrass Cymodocea nodosa 598
in Ria Formosa lagoon (Southern Portugal) after a decade. Cahier de Biologie Marine 599
55:101-108. 600
Cabaço S, Santos R, Duarte CM (2008) The impact of sediment burial and erosion on 601
seagrasses: a review. Estuarine, Coastal and Shelf Science 79 (3):354-366. 602
Carpenter J (2011) May the best analyst win. Science 331:698–699. 603
Costa M, Barrote I, Silva J, Olivé I, Costa MM (2015) Epiphytes modulate Posidonia 604
oceanica photosynthetic production, energetic balance, antioxidant mechanisms and
605
oxidative damage. Frontiers in Marine Science doi.org/10.3389/fmars.2015.00111. 606
Costanza R, d'Arge R, de Groot R, Farber S, Grasso M, Hannon B, Limburg K, Naeem S, 607
O'Neill RV, Paruelo J, Raskin RG, Sutton P, Van den Belt M (1997) The value of the 608
world's ecosystem services and natural capital. Nature 387:253-260. 609
Crain CM, Kroeker K, Halpern BS (2008) Interactive and cumulative effects of multiple 610
human stressors in marine systems. Ecology letters 11:1304-1315. 611
Danovaro R, Fabiano M, Boyer M (1994) Seasonal changes of benthic bacteria in a 612
seagrass bed (Posidonia oceanica) of the Ligurian Sea in relation to origin, 613
composition and fate of the sediment organic matter. Marine Biology 119:489-500. 614
Dattolo E, Ruocco M, Brunet C, Lorenti M, Lauritano C, Sanges R, De Luca P, Procaccini 615
G (2014) Response of the seagrass Posidonia oceanica to different light 616
environments: insight from a combined molecular and photo-physiological study. 617
Marine Environmental Research 101:225-236. 618
Dattolo E, Marín-Guirao L, Ruiz J, Procaccini G (2017) Long-term acclimation to reciprocal 619
light conditions reveals depth-related selection in the marine foundation species 620
Posidonia oceanica. Ecology and Evolution DOI:10.1002/ece3.2731
621
Delgado O, Ruiz J, Perez M, Romero R, Ballesteros E (1999) Effects of fish farming on 622
seagrass (Posidonia oceanica) in a Mediterranean bay: seagrass decline after organic 623
loading cessation. Oceanologica Acta 22:109–117. 624
Erftemeijer PLA, Lewis RRR (2006) Environmental impacts of dredging on seagrasses: a 625
review. Marine Pollution Bulletin 52:1553–1572. 626
Falkenberg LJ, Connell BD, Russel SD (2010) Disrupting the effects of synergies between 627
stressors: improved water quality dampens the effects of future CO2 on a marine 628
habitat. Journal of Applied Ecology 50:51-58. 629
Fonseca MS, Cahalan JA (1992) A preliminary evaluation of wave attenuation by four 630
species of seagrass. Estuarine, Coast Shelf Science 35:565-576. 631
Fonseca MS, Koehl MAR (2006) Flow in Seagrass canopies: the influence of patch width. 632
Estuarine, Coastal and Shelf Science 67:1-9. 633
Fourqurean JW, Robblee MB (1999) Florida Bay: A recent history of ecological changes. 634
Estuaries 22:345-357. 635
Frederiksen MS, Holmer M, Diaz-Almela E, Marba N, Duarte CM (2007) Sulfide invasion in 636
the seagrass Posidonia oceanica at Mediterranean fish farms: assessment using 637
stable sulphur isotopes. Marine Ecology Progress Series 345:93-104. 638
Garcìa-Sanz T, Ruiz-Fernandez JM, Ruiz M, Garcia R, Gonzalez MN, Perez M (2010) An
639
evaluation of a macroalgal bioassay tool for assessing the spatial extent of nutrient
640
release from offshore fish farms. Marine Environmental Research 70:189-200.
641
Garcia R, Holmer M, Duarte CM, Marbà N (2013) Global warming enhances sulphide 642
stress in a key seagrass species (NW Mediterranean). Global Change Biology 643
19:3629–3639. 644
Gera A, Pagès JF, Arthur R, Farina S, Roca G, Romero J, Alcoverro T (2014) The effect of 645
a centenary storm on the long-lived seagrass Posidonia oceanica. Limnology and 646
Oceanography 59:1910-1918. 647
Giakoumi S, Halpern BS, Michel LN, Gobert S, Sini M, Boudouresque CF, Gambi MC, 648
Katsanevakis S, Lejeune P, Montefalcone M, Pergent G, Pergent-Martini C, Sanchez-649
Jerez P, Velimirov B, Vizzini S, Abadie A, Coll M, Guidetti P, Micheli F, Possingham 650
HP (2015) Towards a framework for assessment and management of cumulative 651
human impacts on marine food webs. Conservation Biology 29(4):1228-1234. 652
Gobert S, Sartoretto S, Rico-Raimondino V, Andral B, Chery A, Lejeune P, Boissery P 653
(2009) Assessment of the ecological status of Mediterranean French coastal waters as 654
required by the Water Framework Directive using the Posidonia oceanica Rapid Easy 655
Index, PREI. Marine Pollution Bulletin 58:1727–1733. 656
Govers LL, de Brouwer JHF, Suykerbuyk W, Bouma TJ, Lamersc LPM, Smolders AJP, 657
van Katwijk MM (2014) Toxic effects of increased sediment nutrient and organic matter 658
loading on the seagrass Zostera noltii. Aquatic toxicology 155:253-260. 659
Griffen BD, Belgrad BA, Cannizzo ZJ, Knotts ER, Hancock ER (2016) Rethinking our 660
approach to multiple stressor studies in marine environments. Marine Ecology 661
Progress Series 543:273–281. 662
Gromping U (2006) Relative importance for linear regression in R: The package relaimpo, 663
Journal of Statistical Software 17(1):1–27. 664
Heck KL, Hays C, Orth RJ (2003) A critical evaluation of the nursery role hypothesis for 665
seagrass meadows. Marine Ecology Progress Series 253:123–136. 666
Hemminga MA, Duarte CM (2000) Seagrass ecology. Cambridge University Press 667
Holmer M, Argyrou M, Dalsgaard T, Danovaro R, Diaz-Almela E, Duarte CM, Frederiksen 668
M, Grau A, Karakassis I, Marbà N, Mirto S, Pérez M, Pusceddu A, Tsapakis M (2008) 669
Effects of fish farm waste on Posidonia oceanica meadows: synthesis and provision of 670
monitoring and management tools. Marine Pollution Bulletin 56:1618-1629. 671
Huang D, Ou B, Hampsch-Woodill M, Flanagan JA, Prior RL (2002) High-throughput 672
assay of oxygen radical absorbance capacity (ORAC) using a multichannel liquid 673
handling system coupled with a microplate fluorescence reader in 96-well format. 674
Journal of Agricultural and Food Chemistry 50:4437-4444. 675
Hughes AR, Stachowicz JJ (2004) Genetic diversity enhances the resistance of a 676
seagrass ecosystem to disturbance. Proceedings of the National Academy of 677
Sciences 101:8998–9002. 678
Larkum AWD, Orth RJ, Duarte CM, (2006) Seagrasses: Biology, Ecology and 679
Conservation. Dordrecht (The Netherlands). Springer. 680
Lauritano C, Ruocco M, Dattolo E, Buia MC, Silva J, Santos R, Olivé I, Costa MM, 681
Procaccini G (2015) Response of key stress-related genes of the seagrass Posidonia 682
oceanica in the vicinity of submarine volcanic vents. Biogeosciences 12:4185-4194.
683
Lindeman RH, Merenda PF, Gold RZ (1980) Introduction to bivariate and multivariate 684
analysis, Scott, Foresman, Glenview, Ill. 685
Macreadie PI, Baird ME, Trevathan-Tackett SM, Larkum AWD, Ralph PJ (2014) 686
Quantifying and modelling the carbon sequestration capacity of seagrass meadows – 687
a critical assessment. Marine Pollution Bulletin 83:430-439. 688
Manzanera M, Alcoverro T, Tomas F, Romero J (2011) Response of Posidonia oceanica 689
to burial dynamics. Marine Ecology Progress Series 423:47-56. 690
Marbà N, Duarte CM, Díaz-Almela E, Terrados J, Álvarez E, Martínez R, Santiago R, 691
Gacia E, Grau AM (2005) Direct evidence of imbalanced seagrass (Posidonia 692
oceanica) shoot population dynamics along the Spanish Mediterranean. Estuaries
693
28:53-62. 694
Marbá N, Santiago R, Díaz-Almela E, Álverez E, Duarte CM (2006) Seagrass (Posidonia 695
oceanica) vertical growth as an early indicator of fish farm-derived stress. Estuarine,
696
Coastal and Shelf Science 67:475-483. 697
Marín-Guirao L, Ruiz JM, Dattolo E, Garcia-Munoz R, Procaccini G (2016) Physiological 698
and molecular evidence of differential short-term heat tolerance in Mediterranean 699
seagrasses. Scientific Reports 6:28615. 700
Mazzuca S, Bjork M, Beer S, Felisberto P, Gobert S, Procaccini G, Runcie J, Silva J, 701
Borges AV, Brunet C, Buapet P, Champenois W, Costa MM, D'Esposito D, Gullstrom 702
M, Lejeune P, Lepoint G, Olive I, Rasmusson M, Richir J, Ruocco M, Serra IA, 703
Spadafora A, Santos R (2013) Establishing research strategies, methodologies and 704
technologies to link genomics and proteomics to seagrass productivity, community 705
metabolism, and ecosystem carbon fluxes. Frontiers in Plant Science 4:38. 706
McMahon K, Collier C, Lavery PS (2013) Identifying robust bioindicators of light stress in 707
seagrasses: A meta-analysis. Ecological Indicators 30:7-15. 708
Migliore L, Rotini A, Randazzo D, Albanese NN, Giallongo A (2007) Phenols content and 709
2-D electrophoresis protein pattern: a promising tool to monitor Posidonia meadows 710
health state. BMC Ecology 7:6. 711
Mittler R (2002) Oxidative stress, antioxidants and stress tolerance. Trends in Plant 712
Science 7:405-410. 713
Montefalcone M (2009) Ecosystem health assessment using the Mediterranean sea- grass 714
Posidonia oceanica: a review. Ecological Indicators 9:595-604.
715
Moshe A, Gorovits R, Liu Y, Czosnek H (2016) Tomato plant cell death induced by 716
inhibition of HSP90 is alleviated by Tomato yellow leaf curl virus infection. Molecular 717
Plant Pathology 17:247-260. 718
Munkes B (2005) Eutrophication, phase shift, the delay and the potential return in the 719
Greifswalder Bodden, Baltic Sea. Aquatic Sciences 67:372–381. 720
Nishizawa-Yokoi A, Tainaka H, Yoshida E, Tamoi M, Yabuta Y, Shigeoka S (2010) The 721
26S Proteasome function and Hsp90 activity involved in the regulation of HsfA2 722
expression in Response to oxidative stress. Plant and Cell Physiology 51:486-496. 723
NOAA (National Oceanic and Atmospheric Administration) (2016) Extended reconstructed 724
sea surface temperature (ERSST.v4). National Centers for Environmental Information. 725
Accessed March 2016. www.ncdc.noaa.gov/data-access/marineocean-data/extended-726
reconstructed-sea-surface-temperature-ersst. 727
Oliva S (2012) Seagrass systems under human pressure: responses and assessment 728
tools. PhD Thesis Universitat de Barcelona (Spain). 729
Oliva S, Mascaró O, Llagostera I, Pérez M, Romero J (2012) Selection of metrics based 730
on the seagrass Cymodocea nodosa and development of a biotic index (CYMOX) for 731
assessing ecological status of coastal and transitional waters. Estuarine, Coastal and 732
Shelf Science 114:7-17. 733
Orth RJ, Moore KA (1983) Chesapeake Bay: an unprecedented decline in submerged 734
aquatic vegetation. Science 222:51-53. 735
Orth RJ, Carruthers TJB, Dennison WC, Duarte CM, Fourqurean JW, Heck KLjr, Hughes 736
AR, Kendrick GA, Kenworthy WJ, Olyarnik S, Short FT, Waycott M, Williams SL 737
(2006) A Global Crisis for Seagrass Ecosystems. Bioscience 56:987-996. 738
Pace ML, Carpenter SR, Cole JJ (2015) With and without warning: managing ecosystems 739
in a changing world. Frontiers in Ecology and the Environment 13:460-467. 740
Pergent G, Boudouresque CF, Dumay O, Pergent-Martini C, Wyllie-Echeverria S (2008) 741
Competition between the invasive macrophyte Caulerpa taxifolia and the seagrass 742
Posidonia oceanica: contrasting strategies. BMC Ecology 8:20.
743
Perry CT, Morgan KM (2017) Bleaching drives collapse in reef carbonate budgets and reef 744
growth potential on southern Maldives reefs. Scientific Reports 7:40581 DOI: 745
10.1038/srep40581 746
Piazzi L, Balata D, Ceccherelli G (2016) Epiphyte assemblages of the Mediterranean 747
seagrass Posidonia oceanica: an overview. Marine Ecology 37:3-41. 748
Prado P, Alcoverro T, Romero J (2008) Seasonal response of Posidonia oceanica 749
epiphyte assemblages to nutrient increase. Marine Ecology Progress Series 359:89-750
98. 751
Re R, Pellegrini N, Proteggente A, Pannala A, Yang M, Rice-Evans C (1999) Antioxidant 752
activity applying an improved ABTS radical cation decolorization assay. Free Radical 753
Biology and Medicine 26:1231-1237. 754
Roca G, Alcoverro T, Krause-Jensen D, Balsby TJS, van Katwijk MM, Marbà N, Santos R, 755
Arthur R, Mascaró O, Fernández-Torquemada Y, Pérez M, Duarte CM, Romero J 756
(2016) Response of seagrass indicators to shifts in environmental stressors: a global 757
review and management synthesis. Ecological Indicators 63:310-323. 758
Rotini A, Belmonte A, Barrote I, Micheli C, Peirano A, Santos RO, Silva J, Migliore L 759
(2013) Effectiveness and consistency of a suite of descriptors to assess the ecological 760
status of seagrass meadows (Posidonia oceanica L. Delile). Estuarine, Coastal and 761
Shelf Science 130:252-259. 762
Ruíz JM, Pérez M, Romero J (2001) Effects of fish farm loadings on seagrass (Posidonia 763
oceanica) distribution, growth and photosynthesis. Marine Pollution Bulletin
42(9):749-764
760. 765
Russell BD, Connell SD (2012) Origins and consequences of global and local stressors: 766
incorporating climatic and non-climatic phenomena that buffer or accelerate ecological 767
change. Marine Biology 159(11):2633-2639. 768
Serra IA, Lauritano C, Dattolo E, Puoti A, Nicastro S, Innocenti AM, Procaccini G (2012) 769
Reference genes assessment for the seagrass Posidonia oceanica in different salinity, 770
pH and light conditions. Marine Biology 159:1269-1282. 771
Short FT, Wyllie-Echeverria S (1996) Natural and human-induced disturbance of 772
seagrasses. Environmental Conservation 23:17-27. 773
Tamburello L, Benedetti-Cecchi L, Ghedini G, Bulleri F (2012) Variation in the structure of 774
subtidal landscapes in the NW Mediterranean Sea. Marine Ecology Progress Series 775
457:29-41. 776
Telesca L, Belluscio A, Criscoli A, Ardizzone G, Apostolaki ET, Fraschetti S, Gristina M, 777
Knittweis L, Martin CS, Pergent G, Alagna A, Badalamenti F, Garofalo G, Gerakaris V, 778
Pace ML, Pergent-Martini C, Salomidi M (2015) Seagrass meadows (Posidonia 779
oceanica) distribution and trajectories of change. Scientific Reports 5:12505.
780
Terrados J, Duarte CM, Kamp-Nielsen L, Agawin NSR, Gacia E, Lacap D, Fortes MD, 781
Borum J, Lubanski M, Greve T (1999) Are seagrass growth and survival affected by 782
reducing conditions in the sediment? Aquatic Botany 65:175-197. 783
Underwood AJ (1997) Experiments in ecology: their logical design and interpretation using 784
analysis of variance. Cambridge University Press 785
Unsworth RKF, Collier CJ, Waycott M, McKenzie LJ, Cullen-Unsworth LC (2015) A 786
framework for the resilience of seagrass ecosystems. Marine Pollution Bulletin 100:34-787
46. 788
Unsworth RKF, Cullen LC, Pretty JN, Smith DJ, Bell JJ (2010) Economic and subsistence 789
values of the standing stocks of sea- grass fisheries: potential benefits of no-fishing 790
marine protected area management. Ocean and Coastal Management 53:218–224. 791
Unsworth RKF, Collier CJ, Waycott M, Mckenzie LJ, Cullen-Unsworth LC (2015) A 792
framework for the resilience of seagrass ecosystems. Marine Pollution Bulletin 100:34-793
46. 794
